Mersin Üniversitesi Dil ve Edebiyat Dergisi, MEUDED, 2018; 15 (1), 85-104.
NOUN SEMANTICS AND NUMBER MARKING
IN TURKISH
Emrah Görgülü1
İstanbul Sabahattin Zaim University
Abstract: This work investigates the semantics of nouns and the marking of nominal number in Turkish. Nouns in Turkish behave in a way that is significantly different from their counterparts in languages like English. They appear in their bare form without number specification and are not pluralized when they co-occur with numerals. The questions raised are why nouns behave the way they do and whether their characteristics can be uniformly captured. In previous work, these questions were not addressed since the focus was mostly on the similarities and differences between nouns and adjectives and whether nouns form a lexical category. I show that nouns not only form a lexical category, but their semantic and morpho-syntactic properties indicate that they pattern with set nouns as a nominal subcategory within the typology of nominal subcategories (Rijkhoff, 2002ab, 2008). The analysis also accounts for number discord as well as lack thereof between plural subjects and verbs.
Key words: Noun semantics, bare nouns, number marking, Turkish
1 İstanbul Sabahattin Zaim University, Faculty of Education, Department of English Language Teaching, İstanbul, Turkey, emrah.gorgulu@izu.edu.tr
TÜRKÇEDE AD ANLAMBİLİMİ VE SAYI BELİRLEME
Özet:Bu çalışma, Türkçede adların anlambilimi ve adsal sayı belirlenmesini incelemektedir. Türkçede adlar İngilizce gibi dillerdeki muadillerinden önemli ölçüde farklı davranmaktadırlar. Özellikle, çıplak olarak ve herhangi bir sayı belirlemesi olmadan bulunabilirler ve sayı adları ve niceleyicileri ile birlikte aynı ortalarda bulunduklarında çoğul eki almazlar. Burada sorulan sorular, Türkçede adların neden bu şekilde davrandıkları ve gösterdikleri özelliklerin sistemli bir biçimde açıklanıp açıklanamayacağıdır. Önceki çalışmalarda bu sorulara çok değinilmemiştir çünkü odak noktası genelde adlarla sıfatların benzerlik ve farklılıkları ile adların sözlüksel bir kategori oluşturup oluşturmadığıdır. Bu çalışmada, Türkçede adların yalnızca sözlüksel bir kategori oluşturmakla kalmayıp, anlamsal ve morfolojik-sözdizimsel bakımdan da Rijkhoff'ta (2002ab, 2008) ortaya atılan ve adsal bir alt kategori olan küme adları özelliklerine sahip oldukları gösterilmektedir. Bu açıklama ayrıca Türkçede çoğul özne ve eylem arasında görünen sayı uyumu ve uyumsuzluğunu da izah etmektedir.
Anahtar sözcükler: Ad anlambilimi, çıplak adlar, sayı belirleme, Türkçe
1. INTRODUCTION
It is well-attested that root nouns in Turkish are significantly different from their counterparts in other languages like English and Dutch in that they have what is often referred to as general number (Schroeder, 1999; Corbett, 2000; Bliss, 2004; Acquaviva, 2005; Bale, Gagnon & Khanjian, 2011; inter alia). This means that nouns are not specified for number in terms of singularity or plurality in their bare form. Consider (1).
(1)
a. b.
çocuk kitap
kid/kids book/books
This characteristic of nouns in Turkish is what makes them different from nouns in such languages as English and Dutch where they are
known as singular count nouns.2 Unlike singular nouns in these languages, nouns in Turkish can appear in their bare form inside noun phrases, as shown in (2).3
(2) a.
Ülkü kitap oku-du. Ülkü book read-pa3st 'Ülkü read a book / books'. 'Ülkü did book reading.' b.
Kütüphane-den kitap çal-ın-dı. library-abl book steal-pass-pa3st 'Books have been stolen from the library.' 'Book-stealing took place at the library.
c.
Masa-da kitap var. table-loc book exist.pres 'There is a book / are books on the table.'
The sentences in (2) clearly illustrate that bare NPs are allowed to appear in different structures in the language. In (2a) and (2b), the bare NP ‘kitap’ book is in a verbal sentence and in (2c) the same one occurs in an existential construction. What is important here is that all the NPs are interpreted as as number-neutral.4 That is to say, the referent of the
2 This classification of singular count nouns includes those such as 'dog' and 'chair'
and excludes mass nouns like 'water' and 'furniture'.
3 Abbreviations: ø = null morpheme; 1 = first person; 3 = third person; abl = ablative
case; acc = accusative case; clf = classifier; dat = dative case; fut = future marker; indef = indefinite determiner; lin-sep = linker plus separating element; pass = passive voice; past = past tense; pl = plural marker; pres = present; redup = reduplication; sg = singular;
4 Note that bare NPs include those that appear without any determiner, number
NP is not specified for either singularity or plurality. Note that singular count nouns in English and Dutch cannot appear in their bare form in any type of sentence without leading to ungrammaticality.
Moreover, it is possible for bare NPs to act like a predicate of plural-marked subject NPs, in addition to being a predicate for singular subject NPs in the language. This is shown in (3).
(3) a. Ülkü öğretmen. Ülkü teacher 'Ülkü is a teacher.' b. Ülkü ve Pınar öğretmen(-ler) Ülkü and Pınar teacher(-pl) 'Ülkü and Pınar are teachers.'
In (3b), even though the subject NP refers a plural entity, the predicate does not need to be plural-marked. In that respect, the characteristics of nouns in the language on the one hand, other languages like English, on the other, are significantly different from one another.
On the other hand, in order to specify the number of an entity as singular denoted by an NP, the element that is employed is the indefinite determiner 'bir' that is phonologically the same as the numeral one in the language.
head noun does not co-occur with a determiner or numeral, it carries accusative case marking and the NP is interpreted as definite.
(i) Ülkü kitap oku-du. Ülkü book read-pa3st ‘Ülkü read a book / books.’ (ii) Ülkü kitab-ı oku-du. Ülkü book-acc read-pa3st ‘Ülkü read the book.’
(4)
Note also that when a noun head appears with the plural marker -lAr5, the NP obligatorily refers to more than one entity. Consider the examples in (5). (5) a. b. çocuk-lar kitap-lar kid-pl book-pl ‘kids’ ‘books’
Another important difference between nouns in Turkish and languages like English is that if there is a numeral or a quantifier inside the NP, the head noun does not get plural marking, as exemplified in (6).
(6) a.
iki / on / elli / birkaç çocuk two / ten / fifty / a few kid
'two / ten / fifty / a few kids.'
b.
*iki / on / elli / birkaç çocuk-lar two / ten / fifty / a few kid-pl
'two / ten / a hundred kids.'
(7)
a. b.
ten / a few kids *ten / a few kid
5 The plural marker –lAr appears as either –lar (e.g. kitap-lar) or –ler (e.g. melek-ler)
due to vowel harmony.
a. b.
bir çocuk bir kitap
indef kid indef book
The grammaticality of the examples in (6a), as opposed to ungrammaticality of those in (6b), indicates that the head noun does not get plural marking when it co-occurs with a numeral or a quantifying element. This is in sharp contrast to the English examples in (7a) and (7b).
The data provided above show that nouns in Turkish display certain differences from their counterparts in other languages with respect to their morpho-syntactic and semantic characteristics. More specifically, it was shown that they are number-neutral in their bare form and can appear without any functional elements such as determiners, numerals or quantifiers. In addition, they do not need the presence of plural marking when they appear with number expressing elements like numerals or quantifiers. The question that arises at this point is why nouns behave the way they do in the language. Specifically, how can one account for the morpho-syntactic and semantic characteristics of nouns in a uniform manner? In this paper, I address these issues and propose an account in which I argue for a lexical semantic analysis.
The structure of this paper is as follows: in Section 2, I give an overview of earlier work on nouns and show why they do not fully account for the facts outlined above. In Section 3, I propose a lexical semantic account of nouns in order to capture their morpho-syntactic and semantic properties. In Section 4, I show that the proposed account also accounts for the number discord as well as lack thereof between plural subjects and verbs in the language Section 5 briefly concludes the paper and provides some suggestions for further work.
2. BACKGROUND
2.1.PREVIOUS WORK
It was noted in the earlier studies that nouns in Turkish are transnumeral in that they are semantically neither singular nor plural in their bare form (Schroeder, 1999; Corbett 2000; Acquaviva, 2005; inter alia).6
6 The fact that nouns in Turkish and some other languages are unspecified for number
was termed differently in the literature. Such nouns were labeled as having general
number, or being number-neutral or transnumeral. Even though they seem to capture
the number-neutrality of nouns, these terms still lead to confusion, as will be shown below.
Corbett (2000, p. 14) argues that Turkish type of languages show an opposition general/singular versus plural where the first form does not specify number for the noun on its own. He goes on to say that expressing number is not impossible in these languages; however, it is done when it matters and not obligatorily in languages like English. On the other hand, some early accounts found in Nilsson (1985, p. 26) and Schroeder (1999, p. 46), make the claim that nouns in Turkish denote 'concepts' or 'kind of things' or categories. That is why they do not specify singularity or plurality in their bare form. Note, however, that while this idea seems to be in line with the fact that nouns in Turkish are not specified for number, there are certain characteristics that they do not share with their counterparts that are also categorized as denoting concepts or kinds of things in other languages. For instance, nouns that are often considered to denote concepts or kinds in languages such as Japanese, Chinese, Vietnamese and Thai obligatorily take classifiers when they co-occur with numerals. Their absence would lead to ungrammaticality. The reason for the presence of classifiers in the environment of numerals is that since nouns in languages like Chinese denote concepts or kinds of things, they are argued to be necessary for individuation or a spatial outline (Aikhenvald, 2000; Rijkhoff, 2002a).
In a more recent analysis found in Ketrez (2004), the idea of Turkish being a classifier language is entertained. Working on the different types of plurality in the language, Ketrez argues that Turkish has a fully-fledged classifier phrase as a syntactic category. Also, it has a classifier system associated with the plural marker -lAr in the language. In other words, the plural marker is treated as the head of the classifier phrase in her analysis. However, this line of reasoning would make wrong predictions in terms of language typology. First of all, classifier languages are known to not have plural marking on the noun, especially when there is a numeral in the structure. This, however, is not the case in Turkish. Moreover, the main function of classifiers, as mentioned above, is to individuate the referent of the noun phrase but there is no such requirement in the language. In that sense, Turkish nouns do not actually pair with their counterparts in those languages, as they do not need the obligatory presence of classifiers when modified by numerals. Therefore, any proposal that Turkish nouns should be categorized along with those that require the presence of classifiers would not be so reasonable.
2.2. THE STATUS OF NOUNS
Note that there is also another line of research with a focus on the existence of individual lexical categories in the language. The distributional similarities between nouns and adjectives in Turkish led some researchers such as Grønbech (1936) and Swift (1963) to posit the idea that these two classes must belong to the same category. This is mostly due to the fact that adjectives can act as nouns and carry nominal marking in Turkish. Similarly, Banguoğlu (1986) and Ergin (2001) make the claim that adjectives must be classified as a sub-category of nouns in the language. Based on the observation that nouns and adjectives behave alike in the language, Rijkhoff (2002ab, 2008) argues that Turkish nouns are flexible in the sense that there is no clear distinction between the two classes. Therefore, Rijkhoff categorizes Turkish along with languages such as Quechua and Hurrian and does not include Turkish nouns in his typological classification of noun subcategories. This conclusion, however, does not help to understand the true nature of nouns. The apparent similarities between nouns and adjectives should not prevent one from investigating nouns as a lexical category in the language. Besides, there are certain operational means that were already proposed to distinguish nouns from adjectives in Turkish. For instance, Göksel and Haznedar (2007, pp. 12-13) and Uygun (2007, 2009) note that there are certain distinctions between the two lexical categories. For instance, predicative adjectives and some complex adjectives do not denote entities and most nouns cannot denote properties. More specifically, adjectives in the predicate position always indicate a property and can never denote an entity. In addition, as far as their semantics is concerned, nouns primarily denote entities and not properties. Second, the way adjectives are interpreted is restricted in the sense that they are lexicalized in terms of meaning, as shown below.
(8) a.
Zengin bir adam / topluluk / aile rich indef man / community / family ‘A rich man / community / family’
b.
Bir zengin biz-e yardım et-ti. indef rich we-dat help do-past ‘A rich person helped us.’
The example in (8a) indicates that an adjective like ‘zengin’ rich may modify different nouns such as ‘adam’ man or ‘topluluk’ community. However, the sentence in (8b) shows us that when the adjective ‘zengin’ appears in the absence of a noun and is functioning as one, it obligatorily refers to a human being, indicating that such terms cannot refer to any object that has the property described by the adjective.
Third, although adjectives bear inflectional morphemes, they are actually not inflected for nominal inflection. As there is no overt pronominal form denoting nouns in Turkish (e.g. one in English), and a nouns can be headless, the inflectional markers appearing morphologically on a noun can appear on an adjective where there is no noun head in the construction. Consider the example in (9).
(9)
Ben büyük-ler-i al-acağ-ım. I big-pl-acc take-fut-1sg ‘I will take the big ones.’
What is important in (9) is there is no head noun in the structure. The adjective is inflected for number and case only in the absence of a head noun. That causes the adjective to look like a noun.
Finally, Braun and Haig (2000) propose a diagnostic test in order to identify prototypical adjectives in the language. They argue that prototypical adjectives are compatible with ‘X bir N(oun)’ constructions, as in (10a). In addition, prototypical adjectives are able to appear in reduplication constructions, as in (10b).
(10) a.
büyük bir araba big indef car ‘a big car’
b.
büs-büyük redup-big ‘very big’
The arguments presented above clearly show that nouns and adjectives are not always indistinguishable in Turkish. In fact, the examples indicate that nouns and adjectives dramatically differ from each other in terms of their semantic, morpho-syntactic properties. Based on these facts, one could argue that nouns and adjectives belong to distinct lexical categories in the language. Therefore, it is necessary to understand the actual properties of nouns, to compare them with their counterparts in other languages and then come up with a general theory of noun semantics and number marking in the language. This is important in terms of finding out why Turkish nouns behave the way they do. In the next section, I introduce a theory of noun subcategories based on their semantics and morpho-syntax, proposed by Rijkhoff (2002ab) and then developed in subsequent work (Rijkhoff, 2008; Seifart, 2009ab), and argue that it uniformly captures the facts about nouns in Turkish and across languages.
3. A THEORY OF NOMINAL SUBTYPES
In his seminal work, Rijkhoff (2002a) investigates more than fifty languages and proposes a typology of six noun types according to their morpho-syntactic properties. Rijkhoff argues that a detailed investigation of nouns within and across languages illustrates that first order nouns (i.e. nouns used for discrete objects in the real world) do not appear to share the same morpho-syntax and semantics with regard to quantification. Specifically, languages differ in terms of (i) whether or not first order nouns appear with a plural marker when modified by a numeral (where n > 1), and (ii) if first order nouns directly co-occur with a numeral or whether numerals need to appear with a classifier.7 A cross-linguistic investigation with respect to these two properties leads to the classification of six nominal subtypes including (i) singular object nouns, (ii) set nouns, (iii) sort nouns, (iv) mass nouns, (v) collective nouns, and (vi) general nouns. In the next section, I will introduce the first three of these noun subtypes that are most relevant to the discussion here, and elaborate on their morpho-syntactic and semantic characteristics. I will then address the question of whether nouns in Turkish fit into Rijkhoff’s typological classification. 8
7 See also Wiese (1997) and Acquaviva (2005) for a morpho-syntactically driven
semantic analysis of nouns in various languages.
3.1. NOUN SUBTYPES
3.1.1. SINGULAR OBJECT NOUNS
Singular object nouns denote only singular countable entities. This type of nouns is obligatorily marked with the plural marker when they are modified by a numeral greater than one. In addition, they do not need the presence of classifiers when modified by numerals. This type of nouns is found in typologically different languages such as English, Hittite, Ket, Dutch, West Greenlandic and Tamil, among others. The examples below are from two unrelated languages (Rijkhoff, 2002a pp. 35-36). (11) (Dutch) twee boek-en two book-pl ‘two books’ (12) (Ket) qo’m qim-n ten woman-pl ‘ten women’
The examples in (11) and (12) clearly show that when a singular object noun co-occurs with a numeral in an NP, the plural marker is present obligatorily. This is in fact true for all singular object nouns whenever reference is made to more than one entity. Moreover, when the NP denotes more than one entity, the presence of the plural marker is also needed regardless of whether there is a numeral or not, as in houses and dogs in English. It should also be noted that this type of nouns always takes singular agreement whenever reference is made to singular entities, as shown in (13).
mass nouns like 'silverware', 'milk' and 'freedom' in English, (ii) collective nouns such as 'family', 'team' and 'committee' in English, and (iii) general nouns that co-occur with general classifiers rather than sortal classifiers. General nouns are found in Yucatec Maya. See Rijkhoff (2002ab, 2008) for more details.
(13)
I bought a car / *car.
The morpho-syntactic and semantic properties of singular object nouns suggest that they must be different from nouns in other languages in certain respects. Rijkhoff argues that the main distinction between different nouns is semantic in nature and therefore proposes a lexical semantic account. He goes on to say that nouns are composed of two lexical features, as shown in Table 1.
Table 1. Lexical semantic features of nouns
[Shape] [Homogeneity]
The feature [Shape] indicates whether the noun denotes an entity with a well-defined outline or not. It has a binary value, namely [+Shape] and [-Shape]. The origin of this concept goes back to Hundius & Kölver (1983) and Lucy (1992), who investigated the properties of nouns in Thai and Yucatec Maya, respectively. They argue for the idea that the meaning definitions of nouns in these languages do not involve the notion of 'spatial boundedness' or 'discreteness'. Nouns denoting discrete spatial entities designate properties that are not characterized as having a definite shape in the spatial dimension. So, there is a mismatch in that part of the lexical meaning of nouns does not include the notion 'shape' even though what they denote are inherently discrete in the real world. It is for this reason that numerals need to combine with a classifier in these languages. The basic function of classifiers is then to act like an 'individualizer' since only discrete entities can be counted. Therefore, the feature [-Shape] correlates with the obligatory use of numeral classifiers and [+Shape] correlates with the absence of classifiers. On the other hand, the feature [Homogeneity] indicates whether the noun denotes entities that have portions or members. Following Goodman (1966), Rijkhoff argues that this feature is similar to notions such as 'likepartedness' or 'dissectiveness'. The term 'dissective' is defined as the property of a predicate if that predicate is satisfied by every part of every individual that satisfies it. Basically, nouns like 'flour' and 'oil' define homogeneous entities since they are both cumulative and dissective. For instance, if some flour is added to a
pile of flour, the bigger pile is also referred to as flour. This gives the property of cumulativity. If some flour is removed, the remaining would still be referred to as flour. Therefore, this type of nouns has the feature [+Homogeneity]. On the other hand, singular object nouns like ‘bicycle’ in English, ‘puisi’ seal in West Greenlandic define non-homogeneous entities since one cannot refer to something as a bicycle or a seal if they are more or less than one bicycle or one seal.
Basically, different combinations of these two lexical semantic features output different noun types. For instance, singular object nouns in English and Ket are lexically specified for the features [+Shape, -Homogeneity]. The feature [+Shape] indicates that the property denoted by the noun has a well-defined outline. Thus, nouns occur without classifiers in NPs. The feature [-Homogeneity], on the other hand, indicates that the property being denoted is strictly not cumulative or divisive. In other words, the entity being denoted does not have parts or portions. In the next section, I consider set nouns and argue that Turkish nouns, based on morpho-syntactic characteristics, belong to this subtype of nouns.
3.1.2. SET NOUNS
Set nouns are different from singular object nouns in that they do not denote singular entities. In that sense, set nouns are number neutral and may refer to one entity or more than one entity. When they are modified by a numeral they are not marked with the plural marker. However, just like singular object nouns, they do not need the presence of classifiers when they co-occur with numerals. Consider (15) and (16), taken from Rijkhoff (2002a, pp. 40-41). (14) (Hungarian) két lány two girl ‘two girls’ (15) (Oromo) gala lamaani two camel ‘two camels’
The examples in (14) and (15) show the combination of set nouns with numerals in NPs. However, these languages have a plural marker and its presence is obligatory when reference is strictly made to pluralities. For instance, while the reference of the noun saree ‘dog/dogs’ is number-neutral, the reference of the noun sareellee ‘dogs’ needs to be plural in Oromo. This type of nouns is called set nouns as a set may contain any number of entities including one (i.e. a singleton set) or more than one (i.e. a collective set). These characteristics of set nouns lead Rijkhoff to argue that they are lexically specified for the features [+Shape, ±Homogeneity]. As noted above, the feature [+Shape] shows that the property denoted by the noun has a definite shape or outline. On the other hand, the feature [±Homogeneity] indicates that the property denoted by the noun is not specified for number. In other words, whether the property has portions or parts is not encoded in the lexical specification of the noun itself.
As illustrated in Section 1, nouns in Turkish display morpho-syntactic properties that are quite similar to those of nouns classified as set nouns in the typology of noun subtypes. Basically, Turkish nouns:
(i). are unspecified for number in their bare form,
(ii). do not take classifiers when they co-occur with numerals, (iii). do not get plural marking when modified by numerals.
Based on these facts, it is reasonable to argue that nouns in Turkish are in fact set nouns, exhibiting all the features set nouns cross-linguistically display. Moreover, as we will see in Section 4, this line of analysis accounts for number marking in the verbal domain.
Note also that Rijkhoff (2002ab, 2008) makes a distinction between what he calls 'number marking' that generally applies to singular object nouns in English and 'nominal aspect marking' that applies to set nouns in Turkish. The main difference between the two is that number marking involves a strict singular/plural distinction. Also, plural marking is obligatory with number marking. Nominal aspect markers, on the other hand, restrict the reference to either singulars or plurals. Consider the examples below.
(16) a. bir çocuk indef kid 'a kid' b. çocuk-lar kid-pl 'kids'
The singularization process in (16a) is in fact indicating that the noun designates the property of a singleton set that excluding pluralities. On the other hand, the plural marked NP in (16b) refers to sets with plural entities. Therefore, pluralization should be regarded as restricting the set to plural entities, excluding singulars. The singularization and pluralization above are in fact specifying the number of elements in the set, and not strictly number marking seen in English. This captures the difference between nouns that have the feature [-Homogeneity] and those that have the feature [±Homogeneity] even though both types of nouns are specified for [+Shape].
To sum up, it was shown in this section that the analysis morpho-syntactic and semantic characteristics of nouns in Turkish led to the conclusion that they pattern with what is known as set nouns within Rijkhoff's (2002ab, 2008) broad typology of noun subcategories. In that sense, the account proposed here contributes to the fine classification of nouns based on their meaning as well as structural properties. Nouns in Turkish display the properties of set nouns and their association with functional elements such as the indefinite marker and the plural marker lead to singularization and pluralization respectively. In the following section, I consider sort nouns that are different from both singular object nouns and set nouns in certain respects. The discussion of sort nouns is important here in order to capture the similarities and differences between this particular type and other noun types.
3.1.2. SORT NOUNS
Sort nouns are also known as transnumeral or number-neutral in terms of their number semantics. However, there are significant differences
between this type of nouns and other noun types to which they seem to be quite similar. First, sort nouns do not directly combine with numerals. They need the obligatory presence of a specific class of words known as 'classifiers'. Consider the examples below.
(17) (Thai)
pèt hâa tua
duck five clf:body ‘five ducks’
(18)
thian sìi lêm
candle two clf:long, pointed object ‘two candles’
As shown in (17) and (18), the numeral needs the presence of a classifier and the noun itself is not marked for number. The absence of classifiers in these cases would lead to ungrammaticality. The reason why this is the case is that sort nouns are often considered to be denoting concepts or kinds. Therefore, they cannot be quantified directly. In other words, the lexical specification of this type of nouns is not set for the feature [Shape], and a classifier that provide individuation is necessary for quantification. This type of nouns is lexically specified for the features [-Shape, -Homogeneity] in Rijkhoff’s typology of noun subtypes. This classification provides us with the explanatory power that would otherwise unavailable, since classifying nouns as transnumeral in Turkish is quite problematic and it does not help capture the distinctions between nouns that are generally considered to be transnumeral. In the next section, I look at the number agreement and disagreement issue in the verbal domain which will further provide evidence for the argument that Turkish nouns are set nouns.
4. VERBAL NUMBER (DIS-)AGREEMENT
Another piece of evidence indicating that nouns in Turkish are in fact set nouns comes from the grammatical phenomenon called ‘number
discord’.9 Rijkhoff (2002ab) argues that another distinction between singular object nouns in languages like English and set nouns in languages such as Oromo is the fact that the systematic number discord between a plural NP and a verbal element is observed with set nouns only. Number discord in languages is explained assuming that the verb may agree with the set in which case we have singular verb agreement on the verb or with the individuals in the set in which case we have plural verb agreement. Rijkhoff notes that verb agreement is always with the single set in languages such as Oromo, Georgian and Lango. Consider the examples from Oromo (19).
(19) a.
Gala lamaani sooloo d’ak’-e. camel two market go-3sg.past ‘Two camels went to the market
b.
Nama lamma-a-ti mana jaara.
man two-lin-sep house build.3sg.pres 'Two people build the house.'
In (19a) and (19b), the verbs have singular agreement marker agreeing with the set, hence singular verb agreement. In other words, the pronominal element in the verbal complex agrees with the set and not with individuals. Thus we have singular verb agreement on the verb even though the subject NP refers to multiple entities. A similar phenomenon in Turkish was also noted in various studies (Sezer, 1978; Bamyacı, Häussler & Kabak, 2014, Özyıldız, 2017). A verb may have singular or plural agreement when the subject NP is plural and the referent is a human or humanized entity. This is illustrated below.
(20)
Dört aday bura-dan ayrıl-dı-ø. four candidate here-abl leave-pa3st ‘Four candidates left here.’
(21)
Dört aday bura-dan ayrıl-dı-lar. three candidate here-abl leave-pa3st-pl ‘Four candidates left here.’
The only difference between the two structures above is the fact that whereas the verb in (20) does not have plural agreement, the one in (21) is marked with the pronominal marker, agreeing with the plural subject NP. The consensus in the earlier analyses was that in those cases in which there is no plural agreement marker on the verb, the plural subject is interpreted as a 'collective'. In contrast to that in those cases in which the verb carries the plural agreement marker, the quantity referred to by the plural subject should be interpreted as a group of 'distinct' entities (cf. Dizdaroğlu 1976, p. 68, Sezer 1978 and Gencan 1979, p. 93f).
This line of reasoning is compatible with the account proposed here in that in the former there is no plural agreement marker on the verb even though reference is made to pluralities in the subject NP. If we argue that in those instances in which the verb agrees with the set and not with the individuals, we can account for the collective reading that the subject NP is assigned. On the other hand, in the latter the plural agreement marker on the verb invokes a reading in which the reference is made to a distinct group of entities. This makes sense if we argue that the verb agrees with distinct individuals in the set. Therefore, the number (dis)agreement on the verb supports the argument that Turkish nouns are set nouns.
5. CONCLUSION
In this paper, I argued against the claims that categorize nouns as denoting concepts and kinds and Turkish as a classifier language. In addition, I provided counterevidence for arguments that there is no categorical distinction between nouns and adjectives and the former do not form a lexical category by themselves in the language. I showed several ways in which nouns and adjectives unambiguously differ from each other. Then I made a three-way distinction between nouns in Turkish, English and Chinese based on their morpho-syntactic and semantic properties. This distinction illustrated that earlier work that categorized Turkish nouns with their counterparts in other languages
with respect to number-neutrality or transnumerality did not fully capture the facts. Based on Rijkhoff (2002ab, 2008), I argued that the morpho-syntactic and semantic properties of nouns in Turkish indicate that they should be classified as set nouns in the typology of noun subtypes. I also argued that what was traditionally known as number marking in Turkish needs to be seen as nominal aspect marking, leading to the process of singularization and pluralization in the language. The characteristic of nouns also accounts for the phenomenon called number discord as it takes place with set nouns only. For future work, it is necessary to investigate further characteristics of nouns in order to better understand their nature and compare and contrast them with other noun subtypes.
REFERENCES
Acquaviva, P. (2005). The morphosemantics of transnumeral nouns. Morphology and
Linguistic Typology, Online Proceedings of the Fourth Mediterranean Morphology Meeting, (pp. 21-23).
Aikhenvald, Y. A. (2000). Classifiers. A typology of noun categorization devices. Oxford: Oxford University Press.
Bale, A. Gagnon M. & Khanjian. H. (2011). Cross-linguistic representations of numerals and number marking. N. Li and D. Lutz (eds.), Proceedings of the
Twentieth Semantic Analysis and Linguistic Theory, (pp. 582-598).
Bamyacı, E. Häussler J. & Kabak, B. (2014). The interaction between animacy and number agreement: An experimental investigation. Lingua, 148, 254-277. Banguoğlu, T. (1986). Türkçenin Grameri. Ankara: Dil Tarih Kurumu.
Bliss, H. (2004). The semantics of the bare noun in Turkish. In I. Mezhevich and M. B. Dobrovolsky (eds.) Calgary Papers in Linguistics 25, 1-65.
Corbett, G. (2000). Number. Cambridge: Cambridge University Press. Ergin, M. (2001). Türk dil bilgisi. İstanbul, Bayrak Basım Yayın Tanıtım.
Goodman, N. (1966). The structure of appearance. (Second Edition). Indianapolis: Bobbs-Merrill.
Göksel, A. & Haznedar, B. (2007). Remarks on compounding in Turkish. Part of the MorboComp Project. University of Bologna.
Grønbech, K. (1936). Der Turkische Sprachbau I. Kopenhagen: Levin-Munksgaard. Hundius, H. & Kölver U. (1983). Syntax and semantics of numeral classifiers in Thai.
Studies in Language 7(2), 165-214.
Ketrez, N. (2004). –lAr marked nominals and three types of plurality in Turkish.
Proceedings of Chicago Linguistic Society, 39, 176-192.
Lucy, A. J. (1992). Grammatical categories and cognition: a case study of the
linguistic relativity hypothesis. (Studies in the social and cultural foundations of
language 13). Cambridge: Cambridge University Press.
Nillson, B. (1985). Case marking semantics in Turkish. Department of Linguistics. University of Stockholm.
Özyıldız, D. (2017). Quantification in Turkish. In D. Paperno & E. Keenan (Eds.),
Handbook of Quantifiers in Natural Language. Vol. 2, (pp. 857-937). Springer.
Rijkhoff, J. (2002b). Verbs and nouns from a cross-linguistic perspective. Rivista di
Linguistica, (14)1, 115-147.
Rijkhoff, J. (2008). On flexible and rigid nouns. Studies in Language, 32(3), 727-752. Schroeder, C. (1999). The Turkish nominal phrase in spoken discourse. Turcologica 40.
Wiesbaden: Harratssowitz.
Seifart, F. (2009a). Towards a typology of unitization. Miraña noun classes compared to noun classifiers and singulatives. Manuscript. Max Planck Institute for Evolutionary Anthropology, Leipzig.
Seifart, F. (2009b). Multidimensional typology and Miraña class markers. New Challenges in Typology: Transcending the borders and refining the distinctions. Patience Epps & Alexandre Arkhipov (eds.), Trends in Linguistics Studies and
Monographs 217 (pp. 365-385). Berlin: Mouton de Gruyter.
Sezer, E. (1978). Eylemlerin çoğul öznelere uyumu. Genel Dilbilim Dergisi. Ankara Dilbilim Çevresi Derneği, Ankara, 25-32.
Swift, B. L. (1963). A Reference Grammar of Modern Turkish. Uralic and Altaic Series, v. 19. Indiana University Press: Bloomington.
Uygun, D. (2007). Lexical Categories in Turkish. Ms. Boğaziçi University.
Uygun, D. (2009). A split model for category specification: lexical categories in Turkish. (Unpublished doctoral dissertation). Boğaziçi University, İstanbul. Wiese, H. (1997). Semantics of Nouns and Nominal Number. In ZAS Papers in
